Safeguard
Supply Chain Security•

How Laptops Get Hacked and How to Secure Them

How laptops get hacked: UEFI bootkits, Secure Boot bypasses, Thunderbolt DMA, evil maid attacks, vulnerable drivers and info-stealers, and how to stop them.

Safeguard Security Team
63 min read

Most laptops get hacked through the software people install and the accounts they are signed in to, not through exotic hardware tricks. An info-stealer hidden in a cracked app, a malicious package pulled into a developer's project, or a laptop left unencrypted in a taxi accounts for far more damage than firmware implants do. The deep attacks are real too. Bootkits have bypassed Secure Boot on fully patched Windows 11, firmware implants have survived disk replacement, and a few minutes alone with a laptop's Thunderbolt port has been enough to read an encrypted drive.

This guide covers both ends. It explains what software a laptop runs before the operating system starts, every major way in, real documented cases with CVE numbers, how to detect each attack, and exactly how to defend against it. It ends with checklists for home users, businesses and laptop makers, and a look at the firmware and software supply chain that decides how quickly any of this gets fixed.

It is a defender's guide written from the attacker's point of view. There is no exploit code here, no tool command lines and nothing that helps anyone get into a laptop they do not own.

A note on the law. Testing or accessing a laptop you do not own, or have no written permission to test, is illegal in most countries. Tools and research projects mentioned below are named only as context for authorised testing and for checking your own devices.

This is part of our device security series, alongside How Wi-Fi Gets Hacked and How to Secure It, How Bluetooth Gets Hacked and How to Secure It and How Phones Get Hacked and How to Secure Them. If you want to know how the hardware itself is made, start with our laptop teardown.

Contents

How do laptops get hacked?

Laptops get hacked in five broad ways: the user runs something malicious; a supplier ships a flawed updater, driver or firmware image; an attacker who already has admin rights digs below the operating system into drivers or firmware; someone with physical access tampers with the ports or the boot chain; or the laptop is simply lost with its data readable.

Ranked by how many people they hit, the everyday attacks are:

  1. Info-stealers. Malware that grabs saved passwords, browser session cookies and tokens, usually arriving with pirated software, fake installers or phishing.
  2. Malicious packages and extensions on developer laptops, which steal the keys to source code, CI systems and package registries.
  3. Account takeover from stolen sessions, where a copied session cookie gets the attacker past multi-factor authentication (MFA).
  4. Loss and theft, which turns into a data breach when the disk is not encrypted.
  5. Unpatched software, including the laptop maker's own update tools and drivers.

Ranked by sophistication, the rare attacks are:

  1. UEFI firmware implants and bootkits, such as MoonBounce, CosmicStrand and BlackLotus, which load before the operating system.
  2. Secure Boot bypasses, including signed but vulnerable boot loaders and leaked or test signing keys.
  3. Physical attacks on the boot chain and disk encryption: Thunderbolt direct memory access, TPM bus sniffing and evil maid tampering.
  4. Supply chain compromises of the maker's update channel, such as ShadowHammer.

The rest of this guide takes each one apart.

What software runs on a laptop before the operating system?

A laptop runs a stack of firmware before Windows, macOS or Linux loads. A security processor (Intel CSME or the AMD Secure Processor) starts first, then UEFI firmware from a flash chip on the motherboard, then a boot manager checked by Secure Boot. Separate chips such as the TPM, embedded controller, SSD and Thunderbolt controller run firmware of their own.

That layering is the most important idea in laptop security. Code lower in the stack runs earlier, with more privilege, and survives things that clean up the layers above it, such as reinstalling the operating system.

Layered diagram of laptop software: security processor and microcode, UEFI firmware in SPI flash, boot manager and loader, kernel and drivers, OEM utilities, and apps, with separate chips for the embedded controller, TPM, SSD, Wi-Fi, Thunderbolt and GPU

Figure 1. The software stack inside a laptop. The red layers run before the operating system and outlive a reinstall.

The security processor

Modern Intel laptops contain the Converged Security and Management Engine (CSME), often still called the Management Engine or ME. AMD laptops contain the AMD Secure Processor, often called the Platform Security Processor or PSP. Each is a small, separate processor inside the main chip that starts before the main CPU cores. It handles early boot verification, firmware TPM functions on many systems and, on some business laptops, remote management. It runs its own firmware, which is invisible to the operating system.

UEFI firmware

UEFI (Unified Extensible Firmware Interface) is the modern replacement for the old PC BIOS, and people still use the two names interchangeably. It is stored in an SPI flash chip on the motherboard. It initialises memory and devices, shows the maker's logo, offers the setup menu, and finally hands control to a boot manager. Most laptop makers do not write it from scratch. They license a base from an independent BIOS vendor (IBV) such as AMI, Insyde or Phoenix, combine it with chip vendor reference code, and customise it, often through an original design manufacturer (ODM) that builds the laptop for them.

Secure Boot and the TPM

Secure Boot is a UEFI feature that refuses to run boot code unless it is signed by a trusted key. Microsoft describes it as blocking "untrusted firmware and bootloaders (signed or unsigned)" from starting. A Trusted Platform Module (TPM) is a security chip, or a firmware function inside the security processor, that stores keys and records a measurement (a cryptographic hash) of each boot stage. BitLocker uses those measurements so that the disk key is only released when the boot chain looks unchanged.

Other chips with their own firmware

The embedded controller handles the keyboard, battery and fans. The SSD has its own controller and firmware. Wi-Fi, Bluetooth, Thunderbolt and the GPU each load firmware too. Our laptop teardown walks through where each of these parts and its firmware comes from.

The operating system, drivers and OEM tools

Above the firmware sits the operating system with its kernel drivers, many written by hardware makers rather than the OS vendor. On top of that sit the laptop maker's own utilities: update assistants, support tools and BIOS flash utilities. Several of the worst laptop incidents of the last decade came from exactly this layer.

What is a laptop's attack surface?

A laptop's attack surface is every path an attacker can use to run code on it, read its data or change its firmware. That includes the software users install, the maker's updaters and drivers, the firmware and boot chain, the security processor, fast external ports such as Thunderbolt, and physical access to the machine itself.

Laptop in the centre with ten entry points: phishing and downloads, malicious packages, OEM updaters, vulnerable signed drivers, browser exploits, UEFI firmware and Secure Boot, the security processor, Thunderbolt and DMA ports, evil maid tampering, and loss or theft

Figure 2. Ten ways into a laptop, coloured by what the attacker needs.

It helps to group the entry points by what the attacker needs:

  • The user's help. Phishing, fake installers, cracked software and browser exploits all need someone to open, run or visit something.
  • A supplier. Malicious packages and extensions, compromised update channels and flawed OEM tools arrive through trusted sources.
  • Privileges or a bug. Firmware implants and vulnerable drivers usually need administrator rights first. Their value to an attacker is persistence and stealth, not the first foothold.
  • Physical access. Thunderbolt and DMA attacks, evil maid tampering and TPM sniffing need minutes to hours alone with the laptop. Theft needs only the laptop.

Which laptop attacks matter most?

For most people and companies, the attacks that matter most are info-stealers, stolen sessions and lost laptops, because they are cheap and common. Firmware implants, Secure Boot bypasses and DMA attacks matter most for people with valuable access or secrets, and for laptop makers, whose single mistake reaches every customer.

Matrix of laptop attacks by attacker cost and number of people exposed, with info-stealers and phishing in the cheap, widespread corner and UEFI implants in the expensive, targeted corner

Figure 3. A qualitative map of laptop attacks by cost and reach. Placement is editorial, based on the cases in this article.

The timeline below shows how the field moved. Firmware attacks went from conference demonstrations to commodity malware sold on forums, while the most damaging business incidents increasingly started with one compromised developer laptop.

Timeline from 2017 to 2025 of major laptop firmware, driver and endpoint attacks, from Intel AMT and ME flaws to LoJax, ShadowHammer, Thunderspy, Dell dbutil, MoonBounce, BlackLotus, LogoFAIL, PKfail, the Snowflake campaign and the Shai-Hulud npm worm

Figure 4. Nine years of laptop attacks that changed defences.

What are UEFI and BIOS firmware attacks?

A UEFI firmware attack places malicious code in, or exploits a bug in, the firmware that runs before the operating system. Code there loads first, can tamper with everything after it, and survives reinstalling Windows or even replacing the disk. Removal usually means reflashing the firmware chip or replacing the motherboard.

How it works

Firmware attacks come in two shapes. A firmware implant is malicious code written into the SPI flash chip, either as a new firmware module or as a modification to an existing one. A bootkit lives on the disk's EFI system partition instead and replaces or hijacks the boot manager. Implants are harder to install and harder to remove. Bootkits are easier to deploy but can be blocked by Secure Boot, which is why so much attacker effort goes into bypassing it.

Either way, the attacker's goal is the same: run first, then patch the operating system loader and kernel as they start, so that malware is injected into a running system that looks clean from the inside.

What the attacker needs

Writing to SPI flash usually needs administrator or kernel rights on the running system, plus a platform that does not lock its flash properly, or a vulnerability that bypasses those locks. Some attacks instead abuse the firmware update process, or arrive through a compromised supply chain before the laptop is ever sold.

Real cases

LoJax (2018). ESET described LoJax as the "first UEFI rootkit found in the wild". The attackers used a legitimately signed kernel driver bundled with the RWEverything utility to write a malicious module into the SPI flash. ESET noted the rootkit was "not properly signed", so Secure Boot was the first defence that could have stopped it, and that cleaning it required reflashing the chip "with a clean firmware image" or replacing the motherboard.

MoonBounce (2022). Kaspersky reported in January 2022 that it had found an implant in the SPI flash that had modified an existing firmware component, CORE_DXE, using inline hooks. Earlier implants such as LoJax and MosaicRegressor had added new modules, so this was stealthier. The infection chain ran in memory only and could "persist in the system across disk formatting or replacement". Kaspersky attributed it with medium to high confidence to APT41 or a closely affiliated group.

CosmicStrand (2022). Kaspersky found a firmware rootkit in images for Gigabyte and ASUS motherboards built around the H81 chipset, apparently in use since the end of 2016. It modified a legitimate UEFI driver so it ran at every boot and hooked its way into the Windows kernel. Most victims appeared to be private individuals, and the infection would keep a computer compromised "even if the operating system is reinstalled".

LogoFAIL (2023). Many laptops let the maker, and sometimes the owner, customise the boot logo. CERT/CC's December 2023 note describes how researchers at Binarly found vulnerabilities in the image parsing libraries that draw that logo. Because those parsers run with high privilege during early boot, CERT/CC says an attacker could use them to reach high-privileged UEFI settings, and that a firmware update bundling a malicious image could trigger the flaws. The note lists CVE-2023-39539 (AMI), CVE-2023-40238 (Insyde) and CVE-2023-5058 (Phoenix). It is a pure supply chain story: one bug class in IBV code, inherited by many brands.

Leaked signing keys at MSI (2023). After the Money Message extortion group attacked MSI in March 2023 and leaked data, Binarly reported that the leak contained firmware image signing keys for 57 MSI products and Intel Boot Guard private keys for 116 MSI products. Boot Guard is the hardware check that is supposed to verify firmware before it runs. Binarly warned the leaked keys could let an attacker sign modified firmware that passes that check, on devices with 11th to 13th generation Intel processors. Intel pointed out that Boot Guard OEM keys are generated by the system manufacturer, not Intel. A security feature is only as strong as the place its private key is kept.

Impact

Firmware compromise is close to total. The attacker can disable security tools before they load, steal disk encryption keys as they are released, and reinfect the operating system after every cleanup. For an organisation, the worst part is uncertainty: a laptop that might have firmware malware cannot be trusted until it is reflashed or retired.

How to detect it

Detection is hard, which is the point of these implants. Practical signals are:

  • Firmware scanning. Kaspersky found MoonBounce through logs from its firmware scanner. Several endpoint security products and dedicated firmware security tools can read the SPI flash and compare it against known good images.
  • Measured boot and attestation. The TPM records each boot stage. Remote attestation through device management tools can flag a laptop whose measurements changed unexpectedly.
  • Unexpected BitLocker recovery prompts. Microsoft notes BitLocker enters recovery mode if it "detects changes to the BIOS or UEFI configuration, critical operating system startup files, or the boot configuration". Treat an unexplained recovery prompt as a security event, not just an annoyance.

How to defend against it

  • Keep firmware updated through the maker's official channel, Windows Update or, on Linux, the LVFS and fwupd service that many laptop makers publish to.
  • Keep Secure Boot on. LoJax shows how much it blocks on its own.
  • Set a firmware setup password so boot settings and security features cannot be switched off casually. Microsoft recommends a BIOS password as defence in depth for BitLocker.
  • Buy business laptops that support hardware-verified boot, such as Intel Boot Guard or AMD Hardware Verified Boot, which Microsoft lists among the stronger implementations of Secure Boot.
  • Limit administrator rights. Almost every firmware implant in this section needed them first.
  • If you suspect firmware compromise on a high-value machine, reflash from a known good image with the maker's help or retire the laptop. Reinstalling the operating system is not enough.

For deeper reading, see our guides to analysing UEFI malware on a small budget and to the firmware supply chain.

What is Secure Boot, and how do attackers bypass it?

Secure Boot only lets a laptop start boot code signed by keys in its firmware's allow list (db) and not in its deny list (dbx). Attackers bypass it by bringing an old, signed but vulnerable boot program that was never revoked, by abusing untrusted or leaked signing keys, or by exploiting firmware bugs that run before the check.

The laptop boot chain in six stages with the documented attack that breaks each one, from security processor flaws and leaked Boot Guard keys through LogoFAIL, PKfail, BlackLotus and bitpixie to vulnerable drivers

Figure 5. Each stage verifies the next. Each documented attack below it breaks one link.

How it works

Secure Boot uses four databases stored in firmware:

  • The Platform Key (PK) belongs to the laptop maker and controls changes to the next level.
  • Key Exchange Keys (KEK) may update the two signature databases. Microsoft's KEK is on almost every Windows laptop.
  • db is the allow list: certificates and hashes trusted to sign boot managers, loaders and option ROMs.
  • dbx is the deny list of revoked certificates and hashes.

Every boot binary must chain to a db entry and must not appear in dbx. The weak points are obvious once you see the structure. If the PK is untrusted, everything below it is. If a vulnerable binary is signed by a db certificate and not yet in dbx, Secure Boot will happily run it.

Secure Boot key hierarchy with Platform Key, KEK, db and dbx, and a table of Microsoft Secure Boot certificates that expire in 2026 with their replacements

Figure 6. How Secure Boot decides what may run, and the Microsoft certificates that expire in 2026.

Real case: BlackLotus, CVE-2022-21894 and CVE-2023-24932

In March 2023, ESET published an analysis of BlackLotus, which it called "the first publicly known UEFI bootkit bypassing" UEFI Secure Boot. It ran "on even fully-up-to-date Windows 11 systems with UEFI Secure Boot enabled", was sold on underground forums for $5,000, and could disable BitLocker, HVCI (memory integrity) and Windows Defender.

The bypass used CVE-2022-21894, which Microsoft had fixed in January 2022. The fix did not help, because the vulnerable boot binaries were still validly signed and had not been added to dbx. An attacker with admin rights could simply bring their own old copy. ESET noted that revoking such binaries is slow because a wrong revocation can make machines unbootable.

Microsoft's response was CVE-2023-24932, a staged set of mitigations that adds a new "Windows UEFI CA 2023" certificate to db, updates the boot manager to a 2023-signed version, puts the old "Windows Production PCA 2011" certificate in dbx, and optionally sets a Secure Version Number (SVN) so older boot managers refuse to run. Microsoft warns that once revocations are applied they cannot be reverted, that old recovery and install media will need updating, and that some devices may enter BitLocker recovery. It also says the mitigations are preventive, not a cleanup for already infected machines.

Real case: PKfail, CVE-2024-8105

In July 2024, Binarly reported that many laptop and PC makers had shipped firmware whose Platform Key was a test key generated by AMI and labelled "DO NOT TRUST". BleepingComputer's coverage cites 813 affected products from ten vendors, among them Acer, Dell, Fujitsu, Gigabyte, HP, Intel, Lenovo and Supermicro, with vulnerable firmware dating from May 2012 to June 2024 and 22 unique untrusted keys. A private key for one of these test keys had already leaked. With the private half of a PK, an attacker can rewrite KEK, db and dbx and sign any boot code they like. CERT/CC's note VU#455367 tracks this as CVE-2024-8105 and recommends firmware updates from the vendor, including through LVFS.

Real case: CVE-2024-7344

In January 2025, ESET described a Microsoft-signed UEFI application, reloader.efi, shipped in several system recovery suites such as Howyar SysReturn. It was signed through Microsoft's third-party certificate, Microsoft Corporation UEFI CA 2011, but used its own loader that ran code from an encrypted file "regardless of the UEFI Secure Boot state". An attacker with admin rights could plant it on the EFI system partition. Microsoft revoked the vulnerable binaries in the January 2025 Patch Tuesday dbx update, and ESET saw no exploitation in the wild.

The 2026 certificate expiry

Several of the certificates Secure Boot relies on reached the end of their life this year. Microsoft lists Microsoft Corporation KEK CA 2011 expiring on 24 June 2026, Microsoft UEFI CA 2011 on 27 June 2026, and Microsoft Windows Production PCA 2011 on 19 October 2026, each with a 2023 replacement. Microsoft says devices without the new certificates "will keep starting and running normally", but will stop receiving new boot-level protections, such as dbx revocations and boot manager fixes. A laptop that never received the 2023 certificates is quietly falling behind on exactly the defences this section describes. Our guide to what to check now that Secure Boot certificates have expired has the details.

Impact

A Secure Boot bypass turns a temporary admin compromise into a persistent one that loads before every security tool. It also undermines BitLocker's assumptions, because the boot chain the TPM measures is now under the attacker's control.

How to detect it

  • On Windows, Microsoft's CVE-2023-24932 guidance describes the UEFICA2023Status registry value, which reads "Updated" once the new certificate is in db, and event IDs 1037 and 1042 that confirm the revocation and SVN steps.
  • ESET recommends checking whether known vulnerable binary hashes are present in dbx, and using TPM remote attestation to detect unauthorised boot changes.
  • Watch for unexpected files or changes on the EFI system partition. ESET suggests restricting access to it with file access rules.

How to defend against it

  • Apply Windows updates and firmware updates so the 2023 certificates, dbx revocations and boot manager updates arrive. For managed fleets, follow Microsoft's phased CVE-2023-24932 guidance and test one device per model first.
  • Update recovery and install media after revocations, and back up BitLocker recovery keys before you start.
  • If your laptop was on the PKfail list, install the firmware update that replaces the Platform Key.
  • Do not install boot-level recovery or imaging tools you do not need. CVE-2024-7344 lived in exactly that category.
  • Keep administrator rights scarce. Every bypass above needed them to place files on the EFI system partition.

For background on how these pieces fit together, see Secure Boot, trusted boot and measured boot explained and the software supply chain links inside Secure Boot.

Are Intel ME and AMD PSP a security risk?

Intel's Management Engine and AMD's Platform Security Processor are separate processors inside the main chip with their own firmware and high privileges. Serious vulnerabilities have been found in both, some reachable over the network on business laptops with remote management turned on. They are patched through firmware updates, which many laptops never receive.

How it works

These processors exist to make the platform more secure: they verify early firmware, host firmware TPMs and, on Intel business laptops with vPro, provide Active Management Technology (AMT) for remote administration. The trade-off is that a bug inside them sits below the operating system, cannot be seen by it, and is fixed only by a firmware update that has to travel from the chip vendor through the laptop maker to the user.

A related area is System Management Mode (SMM), a special CPU mode used by firmware that runs with more privilege than the operating system kernel. It is not the security processor, but bugs there have the same flavour: code below the OS that the OS cannot inspect.

Real cases

Intel AMT, CVE-2017-5689 (2017). Intel's advisory INTEL-SA-00075 described how "an unprivileged network attacker could gain system privileges to provisioned Intel manageability SKUs". It was rated Critical, CVSS 9.8. The key word is provisioned: laptops where AMT had been set up for remote management were the ones exposed over the network.

Intel ME, SA-00086 (2017). In November 2017 Intel published a cumulative update for ME, SPS and TXE firmware covering several CVEs, including CVE-2017-5705, buffer overflows in the ME kernel that let an attacker with local access "execute arbitrary code". Intel credited researchers from Positive Technologies. Intel's remedy was firmware from the system manufacturer, plus a detection tool to check whether a system was vulnerable.

Intel CSME, CVE-2019-0090. Intel's advisory INTEL-SA-00213 describes an access control vulnerability in CSME 11.x that "may allow an unauthenticated user to potentially enable escalation of privilege via physical access". Again, Positive Technologies researchers reported it.

AMD fTPM, faulTPM (2023). Researchers from TU Berlin showed that voltage fault injection against the AMD Secure Processor could expose the full internal state of AMD's firmware TPM. Their paper says an attacker with "2-3 hours of physical access" could defeat BitLocker's TPM-only protector. With a PIN, BitLocker kept some protection depending on PIN complexity.

AMD SMM, Sinkclose, CVE-2023-31315 (2024). IOActive researchers presented Sinkclose at DEF CON in August 2024. It affects AMD processors dating back to 2006, requires the attacker to have already compromised the operating system kernel, and lets them gain SMM privileges that can evade antivirus and persist after the operating system is reinstalled. AMD released patches for affected Ryzen, Epyc and Threadripper families.

Impact

A network-reachable bug in a management engine is the worst case: remote control of a laptop without touching the operating system. Most other bugs in this class are escalation paths, valuable to an attacker who already has a foothold and wants to stay hidden.

How to detect it

Use the vendor's detection tools and your device management inventory to record each laptop's ME or PSP firmware version, then compare it against vendor advisories. Intel's SA-00086 notice, for example, names fixed minimum ME versions for each platform generation.

How to defend against it

  • Install firmware and BIOS updates from the laptop maker. ME and PSP fixes arrive bundled with them.
  • If you do not use Intel AMT remote management, do not provision it. If you do, restrict who can reach it on the network.
  • Retire laptops that no longer receive firmware updates from the manufacturer. Intel's SA-00086 advisory notes it would not ship updates for older, unsupported ME versions.
  • For high-value machines with AMD firmware TPMs, add a pre-boot PIN, as discussed in the disk encryption section below.

What are Thunderbolt and DMA attacks, and what was Thunderspy?

Thunderbolt and USB4 carry PCIe, which lets devices read and write system memory directly through direct memory access (DMA). A malicious device plugged into such a port can steal keys or bypass the lock screen in minutes. An IOMMU, used by Windows Kernel DMA Protection and by Macs, limits each device to memory set aside for it.

How it works

DMA is why external GPUs and fast docks work: the device moves data without asking the CPU each time. Microsoft's description of the risk is direct. Drive-by DMA attacks "usually take just a few minutes, with simple-to-moderate attacking tools", for example plugging in a device "while the device owner is on a break" and walking away "with all the secrets on the machine".

The defence is an IOMMU (input/output memory management unit), which gives each device its own map of memory it is allowed to touch.

Two diagrams: without an IOMMU a malicious Thunderbolt device can reach all memory; with an IOMMU and Kernel DMA Protection it can reach only its mapped buffer, and unknown devices wait until the user unlocks

Figure 7. DMA makes Thunderbolt fast. An IOMMU keeps that speed from becoming access to everything.

What the attacker needs

Physical access to a port, a few minutes, and a laptop that is on, asleep or locked rather than shut down. Thunderspy needed a screwdriver too.

Real case: Thunderspy (2020)

Researcher Björn Ruytenberg disclosed Thunderspy in 2020: seven flaws in Intel's Thunderbolt design enabling nine exploitation scenarios. With "5 minutes alone" with a computer, a screwdriver and some portable hardware, an attacker could read and copy data even from an encrypted drive while the laptop was locked or asleep, clone a user's authorised Thunderbolt devices and disable Thunderbolt security. Ruytenberg reported that all Thunderbolt-equipped systems shipped from 2011 to 2020 were vulnerable, that systems with Kernel DMA Protection were only partially vulnerable, and that the flaws "cannot be fixed in software". Macs running macOS were partially affected, and fully affected when running Windows or Linux through Boot Camp.

Impact

The prize is memory: disk encryption keys held after unlock, credentials, and the code that enforces the lock screen. A DMA attack against a sleeping, encrypted laptop can defeat the encryption entirely, because the key is already in memory.

How to detect it

On Windows, open Windows Security, then Device security, then Core isolation details, and check Memory access protection, or look for Kernel DMA Protection set to ON in System Information. DMA attacks themselves leave little trace, so the useful detection is confirming the protection is on.

How to defend against it

  • Turn on Kernel DMA Protection where supported. Microsoft says it is enabled automatically on supporting systems, but it depends on virtualisation for I/O (VT-d on Intel) being on in firmware. When it is on, peripherals with incompatible drivers are blocked if plugged in while the screen is locked or before sign-in.
  • Know its limits. Microsoft states it "only protects against drive-by DMA attacks after the OS is loaded"; firmware has to protect the ports during boot. It also does not cover 1394/FireWire, PCMCIA, CardBus or ExpressCard.
  • Macs use an IOMMU per DMA agent on Apple silicon, so PCIe and Thunderbolt devices "can access only memory that has been explicitly mapped for their use". Intel-based Macs start their IOMMU in a default-deny state.
  • Shut down or hibernate, rather than sleep, when the laptop leaves your control. Ruytenberg and Microsoft both make this point.
  • Disable Thunderbolt in firmware on machines that never use it, as the Thunderspy site suggests; USB-C charging usually still works.
  • Only connect your own Thunderbolt peripherals, and do not leave them unattended either.

Our Bluetooth guide covers the wireless side of peripheral attacks.

What is an evil maid attack, and does disk encryption stop it?

An evil maid attack is any attack by someone with brief, unsupervised physical access to a laptop, such as a hotel cleaner, a colleague or a border desk. Disk encryption stops a thief from reading a powered-off laptop. It does not stop someone who tampers with the laptop, then waits for you to unlock it.

How it works

The name comes from the hotel scenario: you leave the laptop in your room, someone with a key spends a few minutes with it, and you return none the wiser. The attacker has two broad options:

  1. Read the data now, by attacking the disk encryption directly while the laptop is in their hands.
  2. Tamper now, collect later, by planting a keylogger, a bootkit or a modified device so that your next unlock hands them the key.

Disk encryption is the main defence against the first and only a partial defence against the second.

BitLocker's protector options

On Windows, BitLocker can unlock the drive in several ways. Microsoft's countermeasures page lists them:

  • TPM-only, the default, needs no user interaction. The TPM releases the key if the boot chain looks unchanged. Microsoft calls it "more convenient for sign-in but less secure than the other options".
  • TPM with PIN requires a PIN before the TPM releases the key, and the TPM's anti-hammering protection slows PIN guessing.
  • TPM with startup key, and TPM with startup key and PIN, add a USB key.

Microsoft's own guidance for an "attacker with skill and lengthy physical access" is a TPM with an enhanced alphanumeric PIN, standby disabled, and shutting down or hibernating the device "before it leaves the control of an authorized user". It also notes that these settings are not configured by default.

Real case: TPM bus sniffing (2019)

In March 2019, Denis Andzakovic of Pulse Security showed that on a laptop with a discrete TPM chip and BitLocker in its default TPM-only mode, the volume master key could be captured from the LPC bus between the TPM and the processor as the TPM released it during boot, then used to decrypt the drive. The article's conclusion was simple: "Enabling BitLocker with a TPM+PIN protector should mitigate this vulnerability." At a conceptual level, the weakness is that in TPM-only mode the key crosses a physical wire on the motherboard with no secret from the user involved. Firmware TPMs that run inside the processor's security processor have no separate chip and bus to tap, though, as faulTPM showed, they have weaknesses of their own.

Real case: bitpixie, CVE-2023-21563 (2024)

At the 38th Chaos Communication Congress in December 2024, a researcher presented bitpixie, a software attack that bypassed BitLocker on "a fully up-to-date Windows 11 system using Secure Boot". It abuses a Windows Boot Manager flaw, CVE-2023-21563, that Microsoft described as fixed since November 2022, by downgrading to an older, still-trusted boot manager. The talk noted BitLocker "is starting to be enabled by default" on newer Windows 11 installations through Device Encryption, which makes TPM-only the most common configuration. No case needs to be opened.

The structural fix is the CVE-2023-24932 work described earlier: once the old PCA 2011 certificate is in dbx and an SVN is set, older boot managers stop being trusted. A pre-boot PIN defends against the class regardless, because the TPM will not release the key without it.

FileVault on Macs

On Apple silicon and T2 Macs, Apple states that "all FileVault key handling occurs in the Secure Enclave; encryption keys are never directly exposed to the CPU", and that when FileVault is on, the key encryption key is protected by "a combination of the user's password and hardware UID". With FileVault off, the volume is still encrypted, but the key is protected only by the hardware, so anyone who can boot the Mac can read it. Turn FileVault on. Our guide on how to check whether FileVault is on walks through it.

Matrix of six physical attacks against four disk encryption setups, showing TPM plus PIN with hibernate and FileVault on Apple silicon resisting most attacks while TPM-only BitLocker is exposed to bus sniffing and bitpixie

Figure 8. Which disk encryption setup resists which physical attack. Every setup is weaker against someone who tampers and then watches you unlock.

Impact

Against TPM-only BitLocker, a prepared attacker with physical access can read the disk. Against any setup, a successful evil maid can capture your password or PIN on the next unlock and take everything.

How to detect it

  • Tamper-evident measures for high-risk travel: noting screw positions, using tamper-evident seals, keeping the laptop on your person.
  • An unexpected BitLocker recovery prompt, or a laptop that boots differently from usual, after it was out of your sight.
  • Attestation in managed fleets, which flags a changed boot chain.

How to defend against it

  • Turn on full disk encryption everywhere: BitLocker or Device Encryption on Windows, FileVault on Mac, LUKS on Linux.
  • On Windows laptops that carry sensitive data or travel, use TPM plus PIN and hibernate instead of sleep.
  • Apply the Secure Boot revocation updates so downgrade attacks stop working.
  • Keep the laptop with you when you travel. Shut it down at borders and in hotel rooms.
  • Set a firmware password and turn off booting from external and network devices if you never use them.
  • If you have reason to think a laptop was tampered with, stop typing secrets into it and have it examined.

Our posts on choosing between TPM-only and TPM with PIN and on the TPM's own supply chain go deeper.

How do OEM updaters and drivers put laptops at risk?

Laptop makers preinstall update assistants, support tools and the kernel drivers those tools rely on. These run with high privilege, talk to the internet and are trusted by the operating system. When they are flawed or their update servers are compromised, the maker's own software becomes the attack path to millions of laptops at once.

How it works

An OEM updater typically runs as a system service, checks the maker's server for new BIOS, driver and app versions, and installs them. To flash firmware it often ships a kernel driver that can read and write hardware directly. That combination, network access plus kernel-level helper, is exactly what an attacker wants. Three failure modes recur: the driver does not check who is calling it; the updater does not properly verify what it downloads; or the update channel itself is compromised.

Real case: Dell dbutil, CVE-2021-21551

In May 2021, SentinelLabs disclosed flaws in dbutil_2_3.sys, a driver installed by Dell's firmware update tools on computers sold since 2009, which SiliconANGLE described as affecting hundreds of millions of machines. Dell's advisory DSA-2021-088 says the driver "contains an insufficient access control vulnerability which may lead to escalation of privileges, denial of service, or information disclosure", with a CVSS score of 8.8. It came along with BIOS, Thunderbolt, TPM and dock firmware update utilities, and tools such as Dell Command Update, Dell Update, Alienware Update and SupportAssist. Dell's remediation was to update those tools and remove the vulnerable driver. It also warned that an administrator could reinstall the old driver later, and recommended Microsoft's driver block rules. As the next section shows, attackers did exactly that.

Real case: Gigabyte App Center (2023)

In May 2023, Eclypsium reported that firmware on Gigabyte systems dropped a Windows executable during startup, using the Windows Platform Binary Table (WPBT), a firmware feature that lets firmware hand an executable to Windows. That service then downloaded and ran further programs. Eclypsium found one download option used plain HTTP, the HTTPS option did not validate server certificates correctly, and the firmware performed no signature check on what it ran. The affected model list grew from 271 to 406. Gigabyte published updates. Eclypsium's advice was to update firmware, disable the "APP Center Download & Install" setting in firmware setup and set a BIOS password.

Real case: ASUS ShadowHammer (2018 to 2019)

Kaspersky found that in 2018 attackers had compromised ASUS Live Update, the maker's own update tool. The trojanised updater carried a valid ASUS signature and was delivered from ASUS update servers. It was found on more than 57,000 Kaspersky users' machines, but it only activated fully on specific targets: Kaspersky extracted "more than 600 unique MAC addresses" hardcoded into over 200 samples. Kaspersky linked the activity to the ShadowPad actor. Our ShadowHammer write-up covers it in full.

Impact

Each case reached a large installed base through software users never chose and rarely think about. The Dell driver gave local attackers kernel access. The Gigabyte updater gave anyone on the network path a way to run code at boot. ShadowHammer gave a capable attacker signed malware delivered by the vendor itself.

How to detect it

  • Inventory OEM software and drivers across your fleet, including version numbers, and compare them against vendor advisories.
  • Look for the specific artefacts vendors publish, such as the DBUtil_2_3 driver service named in Dell's advisory.
  • Monitor for unexpected new services and scheduled tasks created by OEM tools, and for update tools fetching from unexpected hosts.

How to defend against it

  • Uninstall OEM utilities you do not need. Manage firmware and driver updates centrally instead.
  • Keep the tools you keep up to date, and remove drivers that a vendor advisory says to remove.
  • Turn off firmware features that install software into Windows, such as Gigabyte's App Center option, if you do not want them.
  • Turn on the Microsoft vulnerable driver blocklist, covered next.

What is BYOVD (bring your own vulnerable driver)?

BYOVD is a technique where an attacker who already has administrator rights installs an old, legitimately signed driver with a known vulnerability, then uses that flaw to read and write kernel memory. Windows loads the driver because its signature is valid. The attacker then disables security tools before deploying ransomware or spyware.

How it works

Windows requires kernel drivers to be signed, and Microsoft says malicious actors therefore "are turning to exploit vulnerabilities in legitimate and signed kernel drivers to run malware in kernel". Any signed driver with a flaw that allows arbitrary kernel memory access will do. The attacker does not need the vulnerable driver to already be on the machine; they bring it with them.

Four-stage BYOVD flow from admin rights to dropping a signed vulnerable driver, Windows loading it and the attacker gaining kernel read and write, with the control that stops each stage

Figure 9. The driver is genuine and signed; its bug is what the attacker brings.

What the attacker needs

Administrator rights, a copy of a known vulnerable signed driver, and a system that still allows that driver to load.

Real cases

RobbinHood ransomware and Gigabyte's GDRV.SYS. Sophos documented attackers deploying a legitimate, digitally signed Gigabyte driver with a known vulnerability, CVE-2018-19320, to "(temporarily) disable driver signature enforcement in Windows", load their own unsigned driver and kill security products before running RobbinHood ransomware.

Lazarus and Dell's dbutil driver. ESET reported that in autumn 2021, attackers it attributed to the Lazarus group used Dell's vulnerable driver in a campaign against an employee of an aerospace company in the Netherlands and a political journalist in Belgium, lured by fake job offers. ESET called it "the first recorded abuse of the CVE-2021-21551 vulnerability". The module used the driver's kernel access to switch off seven Windows mechanisms that monitor system activity, which "disables the monitoring of all security solutions on compromised machines".

The community catalogue LOLDrivers lists hundreds of vulnerable and malicious Windows drivers, with detection content for defenders.

Impact

BYOVD is the step that turns an intrusion into a successful ransomware attack: it blinds the tools that would otherwise stop the encryption stage.

How to detect it

  • Alerts on new kernel driver services, especially drivers matching known vulnerable hashes.
  • Code Integrity event logs; Microsoft's documentation describes event 3099 for confirming a driver blocklist policy is active.
  • EDR alerts on tampering with security tools.

How to defend against it

  • Keep the Microsoft vulnerable driver blocklist on. Microsoft says it has been enabled by default for all devices since the Windows 11 2022 update, can be checked in the Windows Security app, and is also enforced when memory integrity (HVCI), Smart App Control or S mode is active. It is updated quarterly and through monthly Windows updates.
  • Turn on memory integrity (HVCI). Microsoft says it is on by default for most new Windows 11 devices.
  • Enable the attack surface reduction rule "Block abuse of exploited vulnerable signed drivers", which stops apps writing such drivers to disk. Microsoft notes it does not block a driver already on the system, which is why the blocklist matters too.
  • Use the latest full blocklist through App Control for Business if you need more than the in-box version, which Microsoft says holds some blocks back for compatibility. Test in audit mode first.
  • Keep admin rights rare. BYOVD always starts with them.

What happens when a laptop is lost or stolen?

A lost or stolen laptop is a data breach unless its disk is encrypted and its accounts can be cut off. With full disk encryption, a powered-off laptop is mostly a hardware loss. Without it, everything on the drive is readable, and saved browser sessions and tokens may give the finder access to cloud accounts too.

How it works

A thief does not need to be a hacker. With an unencrypted drive, they can remove it or boot the laptop from another device and read the files. If the laptop was stolen while unlocked or asleep, they may also have live sessions to email, chat and cloud consoles. Even with encryption, signed-in browser sessions stored on the device remain valid on the server until they are revoked.

Impact

For individuals: identity theft, account takeover and loss of personal files. For organisations: a reportable incident in many jurisdictions. Encryption changes that calculation. Under Article 34 of the GDPR, notifying affected individuals is not required if the controller had applied protection measures, such as encryption, that render the personal data unintelligible to anyone not authorised to access it.

How to defend against it

  • Encrypt every laptop, and make sure the recovery key is stored safely, in your Microsoft or Apple account, your organisation's device management system or a password manager.
  • Turn on device location and remote lock or wipe. On Macs, turning on Find My enables Activation Lock on Apple silicon and T2 models, which requires the owner's Apple Account password before anyone can turn off Find My, erase the Mac or reactivate it. Businesses should use their device management system for remote lock and wipe.
  • Hibernate or shut down in transit, so the laptop is locked at the disk level if taken.
  • Revoke sessions and rotate secrets quickly. Sign out of all sessions for email, cloud and source control accounts, and rotate any tokens or keys stored on the laptop.
  • Keep important data synced or backed up, so the loss is only the hardware.

Our developer laptop offboarding checklist covers the same revocation steps for planned departures.

Why are developer laptops a prime target for info-stealers?

Developer laptops hold the keys to other companies: SSO sessions, cloud credentials, SSH keys, package registry tokens and CI secrets. An info-stealer on one of them can let an attacker publish malicious code, raid a CI system or download customer data. Several of the biggest supply chain incidents of recent years started exactly this way.

How it works

Info-stealers are commodity malware. They arrive through pirated software, fake installers, malicious ads, phishing and, for developers, malicious packages and editor extensions. They copy saved browser passwords, session cookies, cryptocurrency wallets and files that look like credentials, then send them to the attacker within minutes. Stolen session cookies are especially valuable because they represent a login that has already passed MFA.

Scale matters here. Microsoft reported that between 16 March and 16 May 2025 it identified "over 394,000 Windows computers globally infected by the Lumma malware", a single info-stealer family that "steals passwords, credit cards, bank accounts, and cryptocurrency wallets". Stolen credentials are then sold on, sometimes years later.

Flow from infection through what info-stealers take, where it is used and what follows, with four documented cases: CircleCI, LastPass, the Snowflake customer campaign and Shai-Hulud

Figure 10. Info-stealers take what developers are signed in to, not just what is on disk.

Real cases

CircleCI (December 2022). CircleCI's incident report says malware on an engineer's laptop, not detected by antivirus, stole "a valid, 2FA-backed SSO session". The attacker used it to reach production systems and exfiltrate customer environment variables, tokens and keys. CircleCI told customers to rotate every secret stored during the window. Our CircleCI lessons post covers what changed afterwards.

LastPass (2022). LastPass said the attacker in its second incident "targeted a senior DevOps engineer by exploiting vulnerable third-party software". As Sophos summarised, a vulnerable media software package on the engineer's home computer was used to plant a keylogger that captured the employee's master password after MFA, giving access to the engineer's corporate vault and the secrets for cloud backup storage. See our write-up of the home endpoint compromise.

Snowflake customers (2024). Mandiant reported that the attacker it tracks as UNC5537 used Snowflake customer credentials "primarily obtained from multiple infostealer malware campaigns", from families including VIDAR, RISEPRO, REDLINE, RACOON STEALER, LUMMA and METASTEALER. About 165 organisations were notified. The earliest infection behind a used credential dated to November 2020, the impacted accounts lacked MFA, and in several cases the initial infection was on contractor systems "also used for personal activities, including gaming and downloads of pirated software".

Shai-Hulud (September 2025). Krebs on Security reported that a self-replicating worm infected at least 187 npm packages. On developer machines it looked for npm tokens and harvested GitHub tokens, SSH keys and cloud API keys, then used the npm tokens to publish infected versions of other packages the victim maintained. It published the stolen secrets to new public GitHub repositories with "Shai-Hulud" in their names. See our lessons from the Shai-Hulud worm.

Impact

The damage lands far from the laptop: in CI pipelines, package registries, cloud data platforms and every customer of the affected company. That is why a developer laptop deserves stronger controls than most.

How to detect it

  • Sign-in alerts and session activity from unfamiliar locations or devices. CircleCI's investigation began when a customer reported suspicious GitHub OAuth activity.
  • New public repositories, tokens or OAuth apps in developer accounts you did not create, such as the "Shai-Hulud" repositories.
  • EDR alerts on credential store access by unexpected processes.
  • Check exposed credentials against breach and stealer-log monitoring services your security team uses.

How to defend against it

  • Use phishing-resistant MFA everywhere, and require it on data platforms. The Snowflake accounts that were hit had none.
  • Shorten token lifetimes. CircleCI recommends OIDC tokens instead of long-lived stored credentials, and planned periodic automatic token rotation and reduced session trust.
  • Keep secrets off disk. Use a secrets manager, hardware security keys for SSH and signing, and short-lived cloud credentials. Our guide to getting secrets onto developer laptops safely compares the options.
  • Separate work from personal use. No games, pirated software or personal media servers on machines with production access.
  • Control what developers install. Pin dependencies, review new packages and extensions, and use an install-time package firewall.
  • Plan the response. When a developer laptop is compromised, rotate every credential it could reach, not only the ones you know were taken. Our post on developer workstation forensics explains what to collect.

How can you tell if a laptop has been compromised?

Check for changes you did not make: security features switched off, unexpected BitLocker recovery prompts, new drivers or services, unknown sessions and tokens in your accounts, and firmware or boot settings that differ from the maker's defaults. Firmware compromise rarely shows itself, so confirming protections are on matters as much as hunting for signs.

Everyday signs worth checking

  • Security features off. In Windows Security, check that memory integrity, Memory access protection (Kernel DMA Protection) and the vulnerable driver blocklist are on. On a Mac, check FileVault is on.
  • Account activity. New sign-ins, OAuth app grants, API tokens, SSH keys or repositories you did not create.
  • New software, drivers and services you did not install, especially kernel drivers and remote access tools.
  • BitLocker asking for the recovery key after no change you made.

Signs of firmware or boot-level compromise

  • Secure Boot reported as off, or boot order and firmware settings changed.
  • Unexpected files on the EFI system partition.
  • Measured boot or attestation results that differ from the rest of the fleet.
  • Vendor and security tool firmware scans that report a modified image.
  • Firmware versions older than the maker's latest, or missing the 2023 Secure Boot certificates.

If you think you are compromised

  1. Disconnect the laptop from networks, but do not wipe it yet if your organisation will want to investigate.
  2. From a different, trusted device, sign out of all sessions and change passwords for email, password manager, cloud and source control accounts.
  3. Revoke and rotate every token, SSH key and API key that lived on the laptop.
  4. Reinstall the operating system from trusted media. If firmware compromise is suspected, reflash the firmware with the maker's help or retire the device.
  5. Review what the laptop could reach, and look for follow-on activity there.

How do you secure your laptop? A checklist for home users

Install operating system and firmware updates, keep Secure Boot on, turn on full disk encryption with a pre-boot PIN on Windows, hibernate rather than sleep when travelling, set a firmware password, turn on device location and remote wipe, use passkeys, remove unneeded OEM tools and install software only from trusted sources.

Twelve-point laptop security checklist in two columns, from updates and Secure Boot to disk encryption, pre-boot PIN, hibernation, Find My, passkeys, removing OEM tools, the driver blocklist, trusted sources and keeping secrets off disk

Figure 11. The 12-point laptop checklist.

Windows laptops

  1. Turn on automatic Windows updates, and install the laptop maker's BIOS and firmware updates.
  2. Confirm Secure Boot is on, and that memory integrity and the vulnerable driver blocklist are on in Windows Security.
  3. Turn on BitLocker or Device Encryption, and save the recovery key somewhere you can reach without the laptop.
  4. If the laptop holds sensitive data or travels, add a pre-boot PIN and use hibernate instead of sleep.
  5. Check Memory access protection is on if your laptop has Thunderbolt or USB4 ports.
  6. Set a firmware setup password and turn off booting from USB or the network if you never use it.
  7. Uninstall maker utilities you do not need, and keep the rest updated.

Macs

  1. Keep macOS updated, with automatic updates turned on.
  2. Turn on FileVault.
  3. Turn on Find My Mac, which also enables Activation Lock on Apple silicon and T2 Macs.
  4. Leave startup security at its default Full Security setting unless you have a specific need. Apple notes the Reduced Security level allows older macOS versions that have unpatched vulnerabilities.

Linux laptops

  1. Keep Secure Boot on where your distribution supports it.
  2. Use full disk encryption at install time.
  3. Install firmware updates through fwupd and LVFS where your laptop maker publishes them there.

Everyone

  • Use a password manager and passkeys, so stolen passwords and cookies do less damage.
  • Never install pirated software or cracked tools. They are a leading way info-stealers arrive.
  • Keep the laptop with you when you travel, and shut it down at borders and in hotels.

For home network protection, pair this with our router security guide.

How should businesses secure a laptop fleet?

Manage every laptop centrally: enforce encryption with escrowed recovery keys, keep firmware and Secure Boot certificates current, turn on Kernel DMA Protection, the driver blocklist and memory integrity, remove unneeded OEM tools, monitor with EDR and attestation, and give developer laptops stronger controls because their tokens reach production.

Manage the fleet

  • Enrol every laptop in device management, including contractor machines with production access. The Snowflake campaign shows what unmanaged contractor laptops cost.
  • Encrypt and escrow. BitLocker or FileVault on every device, recovery keys stored centrally.
  • Use TPM plus PIN for laptops that hold sensitive data or travel, with hibernation instead of standby, following Microsoft's guidance for skilled attackers with physical access.
  • Manage firmware as carefully as operating system patches. Track BIOS, ME or PSP, and Thunderbolt firmware versions per model.
  • Roll out the 2023 Secure Boot certificates and the CVE-2023-24932 revocations in phases, testing one device per model and updating recovery media.
  • Turn on Kernel DMA Protection, and use the external device enumeration policy to block incompatible devices until sign-in.
  • Enforce the driver blocklist, memory integrity and the BYOVD attack surface reduction rule.
  • Remove OEM utilities you do not need, and push firmware and driver updates from your management tools instead.

A business laptop policy checklist

ControlWhyEvidence from this guide
Full disk encryption with escrowed keysTurns theft into hardware lossGDPR Article 34 encryption provision
TPM plus PIN for high-risk usersStops TPM sniffing and boot manager downgradePulse Security; bitpixie; Microsoft countermeasures
Hibernate, not sleep, in transitKeeps keys out of memoryMicrosoft countermeasures; Thunderspy
Firmware updates tracked per modelFixes ME, PSP, UEFI and Secure Boot keysSA-00086, PKfail, LogoFAIL
2023 Secure Boot certificates and dbxKeeps boot-level protections flowingMicrosoft certificate expiry guidance
Kernel DMA Protection onBlocks drive-by DMAMicrosoft Kernel DMA Protection
Driver blocklist, HVCI, ASR ruleStops BYOVDRobbinHood, Lazarus
Minimal OEM softwareShrinks updater attack surfaceDell dbutil, Gigabyte, ShadowHammer
No admin rights for daily useMost firmware and driver attacks need themLoJax, BlackLotus, CVE-2024-7344
Phishing-resistant MFA and short-lived tokensLimits what stolen sessions can doCircleCI, Snowflake
Developer laptop controlsTheir tokens reach production and registriesLastPass, Shai-Hulud

Developer laptops

Treat developer laptops as part of your software supply chain. Require MFA on every system they touch, prefer OIDC and short-lived credentials in CI, keep signing keys in hardware, review new dependencies and extensions, and make sure your incident plan includes rotating everything a compromised laptop could reach.

Our post on SOC 2 controls for laptops and endpoints maps these controls to audit evidence.

What should laptop makers and firmware suppliers do?

Laptop makers and their firmware suppliers should generate and protect signing keys in hardware security modules, never ship test keys, keep an SBOM for every firmware and driver release, scan what they ship for known vulnerabilities, sign and properly verify every update, remove legacy drivers and installers, and ship protections such as DMA protection on by default.

For laptop makers and ODMs

  • Keys. Generate Platform Keys, Boot Guard keys and firmware signing keys in hardware security modules, unique per product line. Replace every IBV test key before production. PKfail and the MSI leak are what happens otherwise.
  • Inventory. Keep an SBOM for every firmware release and every utility and driver package, so that when an advisory like LogoFAIL lands you know within hours which models are affected.
  • Updates. Sign every update, validate TLS certificates properly, and verify signatures before running anything. Gigabyte's App Center did none of these.
  • Drivers. Enforce access control in every kernel driver, retire old ones, and work with Microsoft to block vulnerable versions. Dell's dbutil driver lived on for years after its replacement shipped.
  • Defaults. Ship Secure Boot, Kernel DMA Protection support, firmware passwords and the 2023 certificates by default, and publish firmware through Windows Update and LVFS so updates actually reach users.
  • Advisories. Publish clear security advisories with affected models and fixed versions, and support firmware for long enough to matter.

For independent BIOS vendors and chip vendors

  • Treat image, file system and network parsers in firmware as attack surface, and test them with fuzzing.
  • Give OEMs a fast, clear way to receive and ship fixes, because every OEM integrates your code differently.
  • Publish SBOMs for reference code and firmware kits so OEMs can track what they inherit.

For software vendors whose tools run on laptops

  • Keep an SBOM for each release of desktop apps, agents and installers, scan the final installer as well as the source, and fix vulnerable components quickly.
  • Never embed secrets in installers, and keep signing keys out of source repositories.

What is inside a laptop's firmware and software supply chain?

A laptop's firmware and software supply chain is every company whose code ends up on it: the chip vendors and their security processor firmware, the independent BIOS vendor, the ODM that customises the firmware, the laptop maker, driver authors, the OS vendor, and every open source component inside all of those. A weakness at any link reaches the laptop.

Almost every case in this guide is a supply chain problem in disguise:

  • LogoFAIL was a bug class in IBV image parsers, shipped by many brands.
  • PKfail was a test key from an IBV that OEMs and ODMs left in production for twelve years.
  • The MSI leak exposed signing keys that should never have been in source code.
  • ShadowHammer was a compromised vendor update channel.
  • Dell dbutil and Gigabyte App Center were vendor drivers and updaters.
  • CVE-2024-7344 was a third-party recovery tool signed through Microsoft's third-party certificate.
  • Shai-Hulud used developer laptops to poison open source packages for everyone else.

Each link has its own owners and release schedule, which is why one unfixed component can sit on millions of laptops. The same pattern runs through every device in this series. Our phone guide shows it in modem firmware and app SDKs, and our post on fabless chip designers and the software supply chain follows it into the silicon. For firmware SBOMs specifically, see SBOMs for firmware and IoT devices.

The practical answer for anyone who builds laptop software or firmware is the same at every layer: know exactly what is in each build, know which components have vulnerabilities that matter, and be able to ship a fix fast.

How Safeguard helps laptop makers and software vendors

Safeguard covers the software supply chain of laptop software in one platform: inventory from source and from shipped binaries such as EXE, MSI and DLL installers and disk images, SBOMs in CycloneDX and SPDX, vulnerabilities ranked by reachability, SAST and secrets scanning on your own code, fix pull requests gated by your CI, Gold patched packages and compliance evidence.

Teams building laptop software usually assemble this from several tools: a source composition analysis (SCA) tool, a separate binary scanner, an SBOM manager, a secrets scanner and a GRC platform for audits. Each keeps its own inventory. Safeguard does the whole job end to end, from the first inventory to the merged fix and the audit evidence, from one inventory and one vulnerability model.

Safeguard flow from source repositories, installers and binaries, extracted firmware images and supplier SBOMs through component identification, SBOM, vulnerabilities and code scanning to fix pull requests, Gold patched packages and evidence

Figure 12. From installer and firmware image to fixed build, in one platform.

Scan the installers and images you actually ship

In the Safeguard console, Integrations > Upload App Binary accepts compiled software including Windows EXE, MSI and DLL files, VM and disk images (QCOW2, VMDK, VHD and OVA), container tarballs, and Android APK and iOS IPA files. That covers OEM utilities and updaters, driver installer packages, desktop agents and the golden images IT teams deploy to laptops. Safeguard unpacks the file, identifies embedded libraries and versions, maps them to known vulnerabilities with EPSS (Exploit Prediction Scoring System) and known exploited enrichment, records licences, and turns the result into a project with a full SBOM. That includes components nobody declared in a manifest.

Firmware images may need to be extracted into their component files first, and binary analysis has limits that apply to any tool: encryption, compression and unusual packing can hide components, and statically linked code may show few identifiable libraries. Pair binary scans with SBOMs from your IBV, ODM and chip suppliers. More detail is in our guide to scanning app binaries with Safeguard and our explainer on binary composition analysis.

SBOMs for every firmware and software release

Every project produces an SBOM you can export in CycloneDX or SPDX, among many other formats. For a laptop maker, that means one SBOM per firmware release, driver package and utility version, ready for an enterprise customer's security review and for the moment the next LogoFAIL-style advisory lands. Attestation and signing with SLSA, Sigstore and in-toto give the SBOM and the build provenance behind it.

Vulnerabilities ranked by what your code actually reaches

A desktop utility or updater with hundreds of open source components can map to hundreds of known vulnerabilities, many in code paths it never calls. Safeguard's reachability analysis ranks vulnerabilities by whether your code can reach them, combined with exploitation signals, so a small platform team works on the handful that matter first.

SCA, SAST and secrets scanning for the code you own

For the code in your repositories, Safeguard runs SCA across the full transitive dependency tree, SAST on your own source, and secrets scanning. SAST catches the class of flaw behind the Gigabyte updater, such as downloads over plain HTTP and TLS certificates that are never validated. Secrets scanning catches private keys and tokens before they are committed, which is exactly how signing keys ended up inside leaked source code in the MSI breach.

Supplier risk and compliance evidence

Safeguard's Third-Party Risk Manager lets you request SBOMs from IBVs, ODMs, driver authors and other suppliers, score them and monitor them continuously. The compliance suite collects evidence for frameworks such as SOC 2 and ISO 27001 from the same scans and fixes, so the work your firmware and platform teams already do shows up in the audit. Safeguard itself is SOC 2 Type I and Type II audited.

Deploy where your code lives

Firmware source and signing infrastructure are among the most sensitive code a company owns. Safeguard runs as SaaS in US and EU regions, in your own VPC, or fully on premises. It can run in Zero AI mode with no language model in the loop, or with your own model.

How this compares

Laptop makers commonly pair a firmware security specialist such as Binarly or Eclypsium, whose research appears throughout this guide and who go deep on UEFI image analysis and firmware integrity, with a source SCA tool such as Snyk or Black Duck and a GRC platform for audits. Those are capable tools. Safeguard's case is consolidation and remediation across the software you build and ship: one inventory across source and the binaries you release, one vulnerability model with reachability, fix pull requests and patched packages rather than tickets, and supplier risk and compliance evidence from the same data. Safeguard can sit alongside a firmware specialist and take over SBOM delivery, open source fixes and audit evidence.

Fix it without waiting for upstream

When a vulnerable open source component sits inside your updater, agent or firmware build, you often cannot wait for its maintainer. Safeguard opens a fix pull request that your own CI builds and tests, and Safeguard Gold supplies patched builds of open source packages when upstream has no fix yet.

This is where Safeguard differs most from a scanner. For code in your repositories, Griffin, Safeguard's remediation model, opens a fix pull request that upgrades or patches the vulnerable dependency. Your CI builds and tests it, and your policies decide what can merge, so nothing changes in your firmware tooling or desktop software without passing your own gates.

For open source components where upstream has not shipped a fix, or the fix is in a version you cannot move to without breaking a product line, Safeguard Gold provides patched, hardened builds of those packages, and Safeguard can fork and patch the vulnerable package so your next build pulls a fixed version. For a laptop maker supporting dozens of models on different code bases, that is the difference between shipping the fix in the next update and carrying a known vulnerability for a product's whole support life.

The firmware stories in this guide show what the alternative looks like. PKfail lasted twelve years. Dell's vulnerable driver was abused months after its fix shipped. Every link that waits for the one above it extends the gap. A team that can patch a vulnerable component in its own build closes its part of that gap in days.

If you build or resell laptops, endpoint software or managed device services and want to offer SBOMs, vulnerability management and fixes to your customers, you can partner with Safeguard.

The takeaway

Laptops are attacked from the top and the bottom. At the top, cheap and constant: info-stealers, phishing, stolen sessions and lost devices. The defences there are equally cheap and are in the checklists above. At the bottom, rare and expensive: firmware implants, Secure Boot bypasses and physical attacks on the boot chain. The answers there are firmware updates, Secure Boot with current certificates, a pre-boot PIN, hibernation and fewer admin rights.

Underneath both is a supply chain. A laptop runs code from chip vendors, BIOS vendors, ODMs, the laptop maker, driver authors, the OS vendor and thousands of open source maintainers, and a developer's laptop is itself a link in other companies' supply chains. The fewer unknown components in that chain, and the faster each known vulnerability is fixed, the less there is for any attacker to use.

Try it yourself

Click through these interactive Safeguard demos to see the steps described in this guide.

Test yourself

How do most laptops get hacked?

Through the software people install and the accounts they are signed in to, such as an info-stealer in a cracked app or a malicious package in a developer's project. Firmware implants are real but far less common.

Why is a UEFI firmware attack so hard to clean up?

Malicious code in the firmware runs before the operating system, so it survives reinstalling the OS or replacing the disk. Removal usually means reflashing the firmware chip or replacing the motherboard, as with LoJax in 2018.

How do attackers bypass Secure Boot?

They bring an old, signed but vulnerable boot program that was never revoked, abuse untrusted or leaked signing keys, or exploit firmware bugs that run before the check. Binarly reported that a 2023 leak from MSI included firmware signing keys for 57 products and Intel Boot Guard keys for 116 products.

What is BYOVD?

Bring your own vulnerable driver. An attacker who already has administrator rights installs an old, legitimately signed driver with a known flaw, and Windows loads it because the signature is valid. They then use it to read and write kernel memory and disable security tools before deploying ransomware or spyware.

Does disk encryption stop an evil maid attack?

It stops a thief from reading a powered-off laptop, but not someone who tampers with the laptop and then waits for you to unlock it. Thunderbolt and DMA attacks matter here too, and an IOMMU limits each device to the memory set aside for it.

Why are developer laptops a prime target for info-stealers?

They hold SSO sessions, cloud credentials, SSH keys, package registry tokens and CI secrets. In December 2022 CircleCI said malware on an engineer's laptop stole a valid, 2FA-backed SSO session, which led to access to production systems.

Name two steps from the home checklist.

Install operating system and firmware updates and keep Secure Boot on. Also turn on full disk encryption with a pre-boot PIN on Windows, set a firmware password, and turn on device location and remote wipe.

Frequently asked questions

Can a laptop be hacked through its firmware?

Yes. Implants such as LoJax, MoonBounce and CosmicStrand were found in laptop and motherboard firmware, and bugs such as LogoFAIL affected firmware from several BIOS vendors. These attacks usually need administrator rights or a compromised supply chain first. Keeping firmware updated, Secure Boot on and admin rights scarce blocks most of them.

Does reinstalling Windows remove a UEFI bootkit?

Not always. A bootkit on the EFI system partition may be removed by a clean reinstall, though ESET noted BlackLotus cleanup also involved removing an enrolled key. A firmware implant in SPI flash survives reinstalls and even disk replacement, as Kaspersky found with MoonBounce. It needs a firmware reflash from the maker or a new motherboard.

Is BitLocker safe without a PIN?

TPM-only BitLocker stops a casual thief reading a powered-off laptop. It is weaker against a prepared attacker with physical access: researchers have sniffed the key from discrete TPM buses, and the bitpixie attack bypassed it in software. Microsoft recommends TPM plus PIN and hibernation for laptops facing skilled attackers.

Is it safe to leave my laptop in a hotel room?

For most people the main risk is theft, so encryption and a strong password are what matter. If you carry sensitive data or are a likely target, shut the laptop down rather than leaving it asleep, use a pre-boot PIN, keep Thunderbolt protections on, and keep the laptop with you where you can. Treat any sign of tampering as an incident.

Should I uninstall my laptop maker's update and support tools?

Remove the ones you do not need. Dell's dbutil driver, Gigabyte's App Center and ASUS Live Update all show that maker tools run with high privilege and have been flawed or hijacked. Keep a way to receive firmware updates, through Windows Update, the maker's current tool or fwupd on Linux, and keep whatever you retain updated.

What should I do if my work laptop is lost or stolen?

Report it to IT immediately so they can lock or wipe it. From another device, sign out of all sessions for email, chat, cloud and source control, and change your passwords. Ask for any tokens, SSH keys or API keys on the laptop to be revoked. If the disk was encrypted and the laptop was off, the data itself is likely safe.

What happens now that Microsoft's 2011 Secure Boot certificates have expired?

Microsoft says laptops without the 2023 certificates keep booting and receiving normal Windows updates, but stop getting new boot-level protections such as dbx revocations and boot manager fixes. Install Windows and firmware updates so the new certificates arrive, and businesses should follow Microsoft's deployment guidance model by model.

How do laptop makers keep vulnerable components out of firmware and drivers?

Keep an SBOM for every firmware, driver and utility release, scan the shipped binaries as well as the source, and track supplier advisories against that inventory. Generate signing keys in hardware security modules and never ship test keys. When a component is vulnerable, patch it in your own build rather than waiting for every supplier upstream.

Never miss an update

Weekly insights on software supply chain security, delivered to your inbox.

We email you a link to confirm, and only then add you to the list. Unsubscribe in one click. Privacy

Self-healing security runs on Safeguard.

Your first fix PR is minutes away.

No sales call required, even your agent can complete the purchase over MCP.